Skip to content

How To Access Iceland Temperature Data

February 22, 2015

By Paul Homewood


Some background and a bit of clarification to the issue of where the Icelandic temperature data comes from.


As I mentioned in the post, The Official Iceland Temperature Series, the Iceland Met Office, IMO, have a homogenised set of temperature data, dating back to 1931. These have been carefully and meticulously adjusted for known changes, such as station moves.

This data is published on their website here.

To access it, you need to scroll down to:




This gives temperature data from 1931 to 2000. No adjustments have been made to this data since 2000, so the rest of the record since then can be accessed from the raw data, here.





This raw data is only given since 1949. However, earlier data can be found in the original Year Books accessed below:






From the left hand side menu, select year and month.




Temperature data is usually on page 2 (the third box). and looks like this. Mean temperatures are in Column 3.




Using this original data, we can see the adjustments which have been made in the homogenised version. For instance, the original mean temperature for Reykjavik in January 1940 was 1.6C.




In the homogenised version, it has been adjusted down to 1.2C.




Smaller changes have been made to other stations, such as Stykkisholmur.


Finally, back in 2012, Trausti wrote an interesting piece about the history of the Reykjavik record.


Measurements in the modern style started in Reykjavík in May 1880. There are earlier observations available but made with unsheltered wall instruments. These can be added to the series back to 1866 and observations were also made in Reykjavík during 1820 to 1854. We will hopefully look at these early measurements later.

The history of the Reykjavík station is quite complicated. The Reykjavík station was operated by the Danish Met. Institute (DMI) from 1880 to 1910 and monthly mean temperatures were published in the yearbooks of the Institute (Meteorologisk Aarbog) which are available online (see an earlier post). During this period there were a few observers and the station was relocated within the town three or four times. A wall-mounted instrument screen was used during the first part of the early period but it was later replaced by a free standing one – the first one in Iceland.

The thermal characteristics of the early screens are better known than elsewhere in Iceland at this time as it was for a considerable period equipped with a thermograph which was read every hour. So we know the diurnal cycle quite well. It conforms well with later-time diurnal cycle observations. A few months of thermograph sheets are missing.

The main thermometer was read three times per day at fixed hours. The DMI always used the average of these + a separate fixed constant for each month (to account for the seasonal variation of the diurnal range). In comparison with modern methods these early calculations tend to have a daytime bias – resulting in slightly elevated temperatures. This is one of the main reasons for the adjustments that have been used later on the means. We will look at this problem later.

The climatic station in Reykjavík stopped making observations in December 1910, temperature measurements continued at a telegraph station in the town but data from this station were never published – and are not used in the attached dataset, except during 1920 and 1921 – which were published.

When the station in Reykjavík was discontinued in 1910 a new one was established just out of town at the newly built Vífilsstaðir sanitarium. This location is slightly warmer in summer but colder in winter than the downtown stations. The published monthly means from this station are included in the attached dataset.

During the early period from 1880 to 1920 a few months are missing from the DMI publication. The Icelandic Meteorological Office (IMO) was established in 1920 and took over the DMI station net. Proper observations began again in Reykjavík (in a wall screen) in May 1920. Observations were made at fixed hours. In 1924 a thermograph was introduced again at the station and during 1924 to 1948 all the published temperature means in Reykjavík were calculated as the average of thermograph readings every two hours. There were daily comparisons of the thermograph and the fixed hour observations.

Since 1949 the Reykjavík monthly mean is calculated as the average of observations made every three hours.

The DMI also published the average monthly minimum temperature during the first few years. A minimum thermometer was, however, in use in Reykjavík during most of the early period, but the means were not published. The minimum thermometer was used for the calibration of the thermograph mentioned above during most of the 1884 to 1907 period.

The IMO has published monthly means of maximum and minimum thermometer readings since 1920. These means are included in the attached dataset.



 The figure shows the annual mean temperature in Reykjavík 1881 to 2011 (as published). No external adjustments have been used.


Relocations are marked with vertical lines. In late 1931 the station was relocated to a rooftop in the town centre and remained there until the end of 1945. The data (above and in the attachment) have not been adjusted for this change nor others. Later versions of the dataset (e.g. the one available at the IMO website) do include adjustments for the relocations. All adjustments are subject to revisions at a later date.

Some internal adjustments are needed during the early part of the series due to later changes in calculation methods. The fixed-hour means that form the basis of both the DMI average method and recent adjustments by the IMO will be made available at this website later.


What is striking is just how methodical the whole process was, even as early as the 1920’s.

  1. February 22, 2015 7:30 pm

    Paul, thanks for this post sharing Trausti’s careful comments. Really highlights the GISS and NCDC homogenization algorithm issue you have been calling attention to.

  2. Lynn Clark permalink
    February 22, 2015 9:18 pm

    One has to wonder how many other national meteorological offices’ carefully-managed climatological data is receiving unnecessary post-hoc adjustments applied by NASA/NCDC, et. al.

    • February 22, 2015 11:03 pm

      KNMI Netherlands, for sure. Essay When Data Isn’t in ebook Blowing Smoke.

  3. February 23, 2015 6:43 am

    “This raw data is only given since 1949. However, earlier data can be found in the original Year Books accessed below:”

    Or you could swallow your pride and just look up GHCN unadjusted. The complete table, back to 1901, is here.

  4. ducdorleans permalink
    February 23, 2015 9:59 am

    please forget my previous excel … didn’t seem correct, and wasn’t … the comparisons that is …

    because if you want to compare GISS to IMO, you have to take into account that GISS calculates average annual as average of spring, summer, autumn and winter, while winter is the average of January, February, and December of the previous year …

    corrected excel sheets are here …

    what are the differences between GISS3 and IMO then ?
    – corrections/homogenizations/whatever from around the 1920/30’s to the 1970’s of up to 1.7°C …
    – absolutely equal from 1970’s onwards …

    what is therefore the point of GISS V3 ?
    “through chartmanship (prof. J. Brignell of numberwatch) and administrative warming (prof. O. Humlun of climate4you) increase the level of catastrophicness by getting rid of the 1940’s blip” …

    hardly any doubt in my mind about that after looking at all that …

Comments are closed.

%d bloggers like this: